AI Benchmark: Running Deep Neural Networks on Android Smartphones

By Andrey D. Ignatov, Radu Timofte, William Chou, Ke Wang, Max Wu, Tim Hartley, Luc Van Gool. ECCV Workshops, 2018

368 citations recorded by the source corpus.

Over the last years, the computational power of mobile devices such as smartphones and tablets has grown dramatically, reaching the level of desktop computers available not long ago. While standard smartphone apps are no longer a problem for them, there is still a group of tasks that can easily challenge even high-end devices, namely running artificial intelligence algorithms. In this paper, we present a study of the current state of deep learning in the Android ecosystem and describe available frameworks, programming models and the limitations of running AI on smartphones. We give an overview of the hardware acceleration resources available on four main mobile chipset platforms: Qualcomm, HiSilicon, MediaTek and Samsung. Additionally, we present the real-world performance results of different mobile SoCs collected with AI Benchmark (http://ai-benchmark.com) that are covering all main existing hardware configurations.

Subfields: Computer Networks and Communications, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/fd4411f035e6ad587d265d01c28522f0e02206c8

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Electrical and Electronic Engineering

Citations in the corpus

Slimmable Neural Networks (2018, background): “These devices have drastically different runtimes for the same neural network (Ignatov et al., 2018), as shown in Table 1.”
Artificial Intelligence in Education: A Review (2020, methodology): “It should be noted that Android Neural Networks API was designed to quickly execute machine learning models on mobile devices [37].”
Deep Joint Source-channel Coding for Wireless Image Transmission (2018, positive): “Inspired by the success of unsupervised deep learning (DL) methods, in particular, the autoencoder architectures [4], [5], we design an end-to-end communication system, where the encoding and decoding functions are parameterized by two convolutional neural networks (CNNs) and the communication chann”
A Simple and Efficient Federated Recommender System (2019, background): “Today’s flagship smartphones can train deep neural networks with millions of parameters [15].”
Convergence of Edge Computing and Deep Learning: A Comprehensive Survey (2019, background): “In addition, particularly for Android smartphones, as one kind of edge devices with mobile CPUs or GPUs, AI Benchmark [54] extensively evaluates DL computation capabilities over various device configurations.”
AutoScale: Energy Efficiency Optimization for Stochastic Edge Inference Using Reinforcement Learning (2020, methodology): “For example, deep learning compiler and programming stacks, such as TVM [12], SNPE [90], and Android NNAPI [4], [51], enable inference execution on diverse hardware back-ends including the co-processors.”